dms dynamics in the most oligotrophic subtropical zones of...
TRANSCRIPT
DMS dynamics in the most oligotrophic subtropical zonesof the global ocean
Sauveur Belviso • Italo Masotti • Alessandro Tagliabue • Laurent Bopp •
P. Brockmann • Cedric Fichot • Guy Caniaux • Louis Prieur •
Josephine Ras • Julia Uitz • Hubert Loisel • David Dessailly •
Severine Alvain • Nobue Kasamatsu • Mitsuo Fukuchi
Received: 17 March 2011 / Accepted: 13 August 2011
� Springer Science+Business Media B.V. 2011
Abstract The influences of physico-chemical and
biological processes on dimethylsulfide (DMS)
dynamics in the most oligotrophic subtropical zones
of the global ocean were investigated. As metrics for
the dynamics of DMS and the so-called ‘summer DMS
paradox’ of elevated summer concentrations when
surface chlorophyll a (Chl) and particulate organic
carbon (POC) levels are lowest, we used the DMS-to-
Chl and DMS-to-POC ratios in the context of three
independent and complementary approaches. Firstly,
field observations of environmental variables (such as
the solar radiation dose, phosphorus limitation of
phytoplankton and bacterial growth) were used along-
side discrete DMS, Chl and POC estimates extracted
from global climatologies (i.e., a ‘station based’
approach). We then used monthly climatological data
for DMS, Chl, and POC averaged over the biogeo-
graphic province wherein a given oligotrophic sub-
tropical zone resides (i.e., a ‘province based’
S. Belviso (&) � I. Masotti � A. Tagliabue �L. Bopp � P. Brockmann
IPSL/Laboratoire des Sciences du Climat et de
l’Environnement, CEA/CNRS/UVSQ, UMR 8212,
CEN Saclay, Bat 701 l’Orme des Merisiers,
91191 Gif-sur-Yvette, France
e-mail: [email protected]
Present Address:A. Tagliabue
Department of Oceanography, University of Cape Town,
Cape Town 7701, South Africa
C. Fichot
Marine Sciences Program, University of South Carolina
Columbia, Columbia, SC 29208, USA
G. Caniaux
CNRM, GAME, Meteo France, CNRS, 42 Avenue G.
Coriolis, 31057 Toulouse Cedex, France
L. Prieur � J. Ras � J. Uitz
Laboratoire d’Oceanographie de Villefranche, Universite
Pierre et Marie Curie, CNRS, UMR 7093, 06238
Villefranche-sur-mer Cedex, France
Present Address:J. Uitz
Marine Physical Lab., Scripps Institution of
Oceanography, University of California San Diego,
9500 Gilman Drive, La Jolla, CA 92093-0238, USA
H. Loisel � D. Dessailly � S. Alvain
Laboratoire d’ Oceanologie et de Geosciences,
Universite Lille Nord de France, UMR 8187,
62930 Wimereux, France
N. Kasamatsu � M. Fukuchi
National Institute of Polar Research,
Research Organization of Information and Systems,
1-9-10 Kaga, Itabashi 173-8515, Japan
Present Address:N. Kasamatsu
Tokyo University of Marine Science and Technology,
4-5-7 Konan, Minato-ku, Tokyo 108-8477, Japan
123
Biogeochemistry
DOI 10.1007/s10533-011-9648-1
approach). Finally we employed sensitivity experi-
ments with a new DMS module coupled to the ocean
general circulation and biogeochemistry model
PISCES to examine the influence of various processes
in governing DMS dynamics in oligotrophic regions
(i.e., a ‘model based’ approach). We find that the
‘station based’ and ‘province based’ approaches yield
markedly different results. Interestingly, the ‘province
based’ approach suggests the presence of a ‘summer
DMS paradox’ in most all of the oligotrophic regions
we studied. In contrast, the ‘station based’ approach
suggests that the ‘summer DMS paradox’ is only
present in the Sargasso Sea and eastern Mediterranean.
Overall, we found the regional differences in the
absolute and relative concentrations of DMS between
5 of the most oligotrophic regions of the world’s
oceans were better accounted for by their nutrient
dynamics (specifically phosphorus limitation) than by
physical factors often invoked, e.g., the solar radiation
dose. Our ‘model based’ experiments suggest that it is
the limitation of phytoplankton/bacterial production
and bacterial consumption of DMS by pervasive
phosphorus limitation that is responsible for the
‘summer DMS paradox’.
Keywords Dimethylsulfide (DMS) � Oligotrophy �Summer DMS paradox � Modelling �Field observations
Introduction
Dimethylsulfide (DMS) is the predominant volatile
sulfur compound in open ocean waters (Andreae 1990)
and an important source of aerosol particles in the
marine atmosphere (Vogt and Liss 2009) and refer-
ences therein; Langley et al. 2010) that could affect
cloud albedo and lifetime. As such, a better knowledge
of the global distribution of DMS in the surface ocean
is required. DMS is a product of dimethylsulfonio-
propionate (DMSP) catabolism which involves differ-
ent enzymatic pathways, not all well understood as yet
(Todd et al. 2007). Since the early nineties, different
approaches have been employed to understand
the global distribution of DMS. The attempts made
to compare seawater DMS concentrations either
with particular oceanic properties (e.g., sea surface
temperature (SST), mixed layer depth (MLD), solar
irradiance, nutrients, chlorophyll a (Chl) or DMSP
concentrations) or combinations of several variables,
have met with mixed success (Andreae 1990; Kettle
et al. 1999; Lana et al. 2011). Although the seasonality
in DMS is certainly determined by the interplay
between numerous physical and biogeochemical pro-
cesses and parameters (Stefels et al. 2007), the most
recent correlation analysis shows that the monthly
ocean surface DMS distribution follows the average
solar radiation intensity experienced by phytoplankton
in the mixed layer (i.e., the surface radiation dose
(SRD)) more closely than it follows SST, Chl or
nutrient concentrations (Lana et al. (2011) after
Vallina and Simo (2007)). Re-examination of the
DMS-SRD relationship at a global scale by Dere-
vianko et al. (2009), seems however to indicate that
when using minimal aggregation methods, SRD
accounts for only a small fraction (4–14%) of the
total variance in DMS concentrations. At local scales,
data show that DMS and SRD can be strongly
correlated (Vallina and Simo 2007), but slopes and
intercepts display little consistency in the Atlantic
Ocean and adjacent seas (Archer et al. 2009; Belviso
and Caniaux 2009; Miles et al. 2009). As shown by
Belviso and Caniaux (2009), part of the discrepancy
arises from the way in which SRDs are calculated. So,
in order to investigate the spatial variations in DMS/
SRD slopes, it is crucial to calculate SRDs in an
identical fashion.
In the Sargasso Sea and in the north-western
Mediterranean Sea, DMS concentrations and SRDs
peak in summer (Dacey et al. 1998; Vallina and Simo
2007). On the other hand, in the north-eastern Atlantic
Ocean off Portugal and in the western English
Channel, DMS levels are markedly higher in spring
than in summer (Archer et al. 2009; Belviso and
Caniaux 2009). Since DMS concentrations and SRDs
have not been monitored concomitantly apart from in
the Atlantic and the Mediterranean Sea, there is not
sufficient experimental evidence to demonstrate
unequivocally that variations in biological factors
are small enough for SRD to be dominant at local
scales (as suggested by Derevianko et al. 2009). Based
on the high correlation between DMS and SRD in the
Sargasso Sea, it is expected that other oligotrophic
subtropical zones of the global ocean would display
similarly high DMS surface levels during the summer
months. The cores of subtropical anticyclonic gyres
are characterized by their oligotrophic status and
minimal Chl concentration, compared to that of the
Biogeochemistry
123
world ocean (Morel et al. 2010 and references therein).
Not only the low Chl level, but also a reduced amount
of colored dissolved organic matter (CDOM) account
for the extreme clarity of waters. Although they may
be similar, different oligotrophic zones are not iden-
tical regarding their Chl and CDOM contents, or their
seasonal cycles (Morel et al. 2010). In spite of its
low Chl, the Sargasso Sea presents the highest
CDOM content amongst the six zones studied by
these authors.
The existence of elevated summer DMS concen-
trations months after surface Chl levels have peaked
has been termed as the ‘summer DMS paradox’ (Simo
and Pedros-Alio 1999). Here we use 3 different
approaches to explore similarities and differences in
terms of ‘summer DMS paradox’ between oligo-
trophic subtropical zones. One is a ‘station based’
analysis. To investigate whether the temporal vari-
ability in DMS concentration is explained by changes
in SRD at some location, while its spatial variability is
controlled by other factors, we assembled a database
of SRDs, sea surface DMS concentrations and
biogeochemical variables at or in the vicinity of the
oligotrophic zones (Table 1) selected by Morel and
Gentili (2009) and Morel et al. (2010). Next comes a
‘province based’ analysis using climatology products
averaged over biogeographic provinces wherein oli-
gotrophic regions reside. Finally a ‘model based’
approach is also used to examine the influence of
various processes in governing DMS dynamics in
oligotrophic regions.
Data and methods
The ‘station based’ approach
Overview of the selected geographic zones
We considered the six quadrangular domains selected
by Morel et al. (2010), and used for comparison a
mesotrophic site off Portugal (site P, see Table 1 and
Fig. 1) which exhibits seasonal oligotrophy in summer
as area S-ext. (see below), as well as a sub-zone in the
oriental basin of the Mediterranean Sea coinciding
with the most oligotrophic waters (Morel and Gentili
Table 1 Geographical information for the ten zones selected
for the present study (see also map in Fig. 1); in addition are
included the Longhurst’s division of the oceans into static
biogeographic provinces, as well as the number of DMS data in
each location (data taken from the Global Surface Seawater
DMS Database)
Location Notation Longitude Latitude Biogeographic provinces
Acronym and reference numberaNumber of
DMS data
N-Hemisphere
Hawaii Islands zone Hb -170 -150.9 10 18 NPTE (38) 232
Mariana Islands zone Mb 150 165.9 10 20 NPTW (35) and WARM (41) 7
South-Sargasso Sea Sb -70 -45 22 27 NASW (6) and NATR (7) 48
North- and South-Sargasso Sea S-ext.c -66 -52 22 34 NASW (6) and NATR (7) 390
Eastern Mediterranean Sea Mdc 33 36 14 32 MEDI (16) 135
Atlantic off Portugal Pb -25 -15 34 40 NASE (18) 241
S-Hemisphere
South Indian gyre Ib 70 90 -30 -21 ISSG (23) 0
Brasilian Atlantic gyre Bb -32 -25.9 -22.5 -12.5 SATL (10) 157
Easter Island zone Eb -125 -100 -30 -20 SPSG (37) 25
Easter Island extended zone E-ext.c -150 -100 -30 -20 SPSG (37) 192
The locations of the PROSOPE, POMME and BIOSOPE cruises, as well as the sampling stations BATS and Hydrostation S, are in
areas denoted Md, P, E and S-ext., respectively
NPTE N. Pacific Tropical Gyre (East); NPTW N. Pacific Tropical Gyre (West); WARM W. Pacific Warm Pool; NASW N. Atlantic
Subtropical Gyre (West); NATR N. Atlantic Tropical Gyre; MEDI Mediterranean Sea, Black Sea; NASE N. Atlantic Subtropical
Gyre(East); ISSG Indian S. Subtropical Gyre; SATL South Atlantic Gyre; SPSG S. Pacific Subtropical Gyrea Longhurst (1998) and Lana et al. (2011)b Morel et al. (2010)c This work
Biogeochemistry
123
2009; denoted Md in Fig. 1). The latter was also
spatially homogeneous in terms of CDOM absorption
of UV radiation at 320 nm (the influence of the Nile
river was discarded as well as that of other coastal
waters). Unfortunately, there are few DMS measure-
ments in the deep blue waters characterizing area S
(Table 1), whereas the BATS station, located in the
northern Sargasso Sea (Fig. 1), is a reference site for
DMS experimentalists. That is why we selected a new
quadrangular domain encompassing the BATS station
(Fig. 1, white dotted contours). It is denoted S-ext. in
Table 1 where ext. stands for ‘extended area’. A
similar approach was applied to the area E located in
the South Pacific gyre in the vicinity of Easter Island
(Table 1; Fig. 1). The new area (E-ext.) extends in
longitude between 100�W and 150�W.
Similarities and differences in terms of climatolog-
ical annual cycles of Chl and POC concentrations
between Morel et al. (2010)’s areas and the two new
ones are evaluated in Fig. 2. The seasonality in Chl and
POC concentrations is not significantly affected by the
longitudinal extension of area E. In the Atlantic Ocean,
conversely, the latitudinal extension and longitudinal
restriction strongly affect both variables. Area S-ext.
does not exhibit permanent oligotrophy. However, the
differences between the two areas are much less
marked during the summer months (J, Ju, A, and S)
than in winter. In other words, both areas S and S-ext.
exhibit spatial and temporal homogeneity in terms of
oligotrophic conditions during the summer period.
DMS, Chl and POC concentrations
We used discrete DMS measurements taken from the
Global Surface Seawater DMS Database (http://saga.
pmel.noaa.gov/dms/) and calculated monthly averages
in each selected zone. Areas M and I are free of DMS
data (Fig. 1). However, some DMS measurements
were carried out west of the south-west corner of area
M (Fig. 1). We arbitrarily allocated these measure-
ments to area M. Chl and POC estimates were
extracted from global climatologies (Table 2).
Physical and biogeochemical variables
We consider both the environmental and biogeo-
chemical parameters that are likely to affect the
production and removal of DMS, and the recently
Fig. 1 Annual mean values of Chl concentration (1997–2010,
in mg m-3) for the global ocean. Following the procedure of
Morel and Gentili (2009), SeaWiFS data were corrected for the
presence of yellow substance (CDOM) in varying proportions
with respect to Chl. The solid white boxes superimposed on the
map show the six oligotrophic zones selected by Morel et al.
(2010), i.e. S, H, E, B, I and M (see also Table 1 for the exact
locations and meaning of the identifiers). The mesotrophic zone
off Portugal (P) is selected for a comparison as in Morel et al.
(2010). The boxes with dotted-white contours (i.e., S-ext., E-ext.
and Md) were selected either to augment the number of DMS
data identified with red dots in each box (S-ext. and E-ext.) or to
exclude the coastal waters from the analysis in the narrow
oriental basin of the Mediterranean Sea (Md). Also shown with
black dashed lines are the ten Longhurst’s static biogeographic
provinces wherein a given oligotrophic subtropical zone resides.
The meaning of the numbers is provided in Table 1
Biogeochemistry
123
proposed notion of mixed layer solar radiation dose
(SRD).
Standardization procedure for SRD estimates In
order to evaluate the average solar radiation intensity
experienced by phytoplankton in the mixed layer,
Vallina and Simo (2007) defined the daily-average
solar radiation dose (SRD) by:
SRD ¼ 1
MLD
ZMLD
0
IðzÞdz ð1Þ
where MLD is the mixed layer depth (units in m) and
I(z) the irradiance (W m-2) at depth z. The MLD is
defined as the depth at which the temperature is 0.1�C
lower than that at 5 m depth (Vallina and Simo 2007;
Miles et al. 2009; Derevianko et al. 2009). For the
decay of the solar irradiance with depth, Vallina and
Simo (2007) considered an exponential law of the
Beer–Lambert form:
IðzÞ ¼ I0 expð�kzÞ ð2Þ
where k is the diffuse attenuation coefficient (units
m-1) and I0 the incident solar irradiance. Only one
exponential was used in Eq. 2 because the authors
were right in what they were assuming that DMS
production was not dependent on the infrared radiation
but on the visible part of the solar radiation available
for phytoplankton growth. However, it is well estab-
lished in the literature that DMS production and
consumption by phytoplankton and bacteria also
respond to UV radiation dose (e.g., Sunda et al.
2002; Toole and Siegel 2004). Here both PAR and
UV doses will be taken into account but treated
independently.
Standardization procedure for SRD estimates in the
PAR domain. We believe that the intensity of SRD
determined by Vallina and Simo (2007) and several
authors (Miles et al. 2009; Derevianko et al. 2009;
Belviso and Caniaux 2009) is largely over-estimated
(Tables 3, 4). This can lead to ambiguous comparisons
when applied at different locations of the global ocean
and finally to divergent analyses when regressed with
DMS measurements. This overestimation is due to two
factors: (i) I0 is generally measured above the sea
surface by pyranometers covering the global solar
radiation spectrum (300–3000 nm) instead of the
photosynthetically available radiation (PAR,
300–700 nm), which constitutes only 40% of the total
irradiance at the Earth’s surface (Jerlov 1974; Jitts
et al. 1976); (ii) I0 has to take account the reflection at
sea surface. Classical reflection coefficient values of
0.06 are reported (Payne 1972; Katsaros et al. 1985)
and used here.
For these reasons, we use the following expression
for computing the SRD:
SRD ¼ 1
MLD
ZMLD
0
ð1� aÞI # expð�kzÞdz ð3Þ
where a is the reflection coefficient (0.06) and I; the
PAR (i.e. 43% the total solar irradiance). In agreement
10
20
30
40
50
NH
SH
Ja
Ju
F
A
M
S
A
O
M
N
J
D
Ju
Ja
A
F
S
M
O
A
N
M
D
J
PO
C (
mg
m-3
)
0
0.02
0.04
0.06
0.08
SS-ext.EE-ext.
Chl
(m
g m
-3)
(a)
(b)
Fig. 2 Climatological annual cycles of averaged a Chl and
b POC concentrations within the zones S, S-ext., E and E-ext.
(see Table 1 for the exact locations and meaning of the
identifiers). The month scale begins with January for the
Northern Hemisphere and in July for the Southern Hemisphere
Biogeochemistry
123
Ta
ble
2S
imil
arit
ies
and
dif
fere
nce
sin
term
so
fp
hy
sica
l,ch
emic
alan
db
iolo
gic
alv
aria
ble
sb
etw
een
the
dif
fere
nt
oli
go
tro
ph
icsu
btr
op
ical
zon
esin
ves
tig
ated
du
rin
gsu
mm
er
Var
iable
sD
ata
sourc
esan
dre
fere
nce
sL
oca
tion
and
month
sF
igure
num
ber
this
wo
rkM
dJu
,S
P A,
SE
-ex
t.B F
,M
D,
JaF
,M
DM
SD
MS
dat
abas
e:h
ttp
://s
aga.
pm
el.n
oaa
.go
v/d
ms/
*-
--
--
--
-5a
DM
S/C
hl
DM
Sd
atab
ase:
htt
p:/
/sag
a.p
mel
.no
aa.g
ov/d
ms/
Corr
ecte
dC
hl
clim
atolo
gy
(More
let
al.
20
10)
*-
--
-*
--
5b
DM
S/P
OC
DM
Sd
atab
ase:
htt
p:/
/sag
a.p
mel
.no
aa.g
ov/d
ms/
PO
Ccl
imat
olo
gy
(Du
fore
t-G
auri
eret
al.
20
10)
*-
--
-*
--
5c
Ven
tila
tion
toat
mo
sph
ere
(DM
Sse
a-to
-air
tran
sfer
vel
oci
ty:
kg
)S
ST
and
win
dsp
eed
clim
ato
log
ies:
htt
p:/
/jis
ao.w
ashin
gto
n.e
du/d
ata/
coad
s_cl
imat
olo
gie
s/
Sc
nu
mb
er:
Sal
tzm
anet
al.
(19
93),
kg
:W
ann
ink
ho
f(1
99
2)
**
**
*6b
Su
rfac
ew
ater
stra
tifi
cati
on
ML
Dcl
imat
olo
gy
(Kar
aet
al.
20
03)
**
--
--
--
6a
UV
ph
oto
nab
sorp
tion
rate
Ph
oto
chem
ical
po
ten
tial
ML
Dcl
imat
olo
gy
(Kar
aet
al.
20
03)
UV
do
wn
war
dir
rad
ian
ce(R
ugg
aber
etal
.1
99
4)
Dif
fuse
atte
nuat
ion
coef
fici
ent
(Fic
hot
etal
.2
00
8)
Ab
sorp
tio
nco
effi
cien
to
fC
DO
M(F
ich
ot
and
Mil
ler
20
10)
**
**
*6d
Ph
yto
pla
nk
ton
gro
up
do
min
ance
aP
HY
SA
T(A
lvai
net
al.
20
08)
PR
Oo
rN
AN
OP
RO
or
NA
NO
SY
Nf
SY
Nf
SY
N8
Ph
yto
pla
nk
ton
pig
men
t-b
ased
size
clas
sesb
Pig
men
td
atab
ase
(Uit
zet
al.
20
06)
PIC
O&
NA
NO
PIC
OP
ICO
nd
nd
7c–
e
Mix
edla
yer
UV
do
se(U
Vst
ress
)M
LD
clim
ato
log
y(K
ara
etal
.2
00
3)
UV
do
wn
war
dir
rad
ian
ce(R
ugg
aber
etal
.1
99
4)
Dif
fuse
atte
nuat
ion
coef
fici
ent
(Fic
hot
etal
.2
00
8)
**
**
*6c
Photo
pro
tect
ion
index
Pig
men
tdat
abas
e(U
itz
etal
.2
00
6)
*-
-*
nd
nd
7b
Lim
itin
gn
utr
ien
tfo
rp
hy
to.
and
bac
teri
alg
row
thD
IPco
nce
ntr
atio
nc
(Mat
her
etal
.2
00
8;
Mo
uti
net
al.
20
08)
*?
??
??
??
?_
DIP
turn
ov
erti
med
(Mo
uti
net
al.
20
08)
--
??
??
nd
nd
Lim
itin
gn
utr
ien
te(M
ou
tin
etal
.2
00
8;
Van
Wam
bek
eet
al.
20
02,
20
08)
Pn
dN
nd
nd
Th
ere
fere
nce
site
isS
-ex
t.S
um
mer
isJ,
Ju,
A,
Sin
the
No
rth
ern
Hem
isp
her
ean
dD
,Ja
,F
,M
isth
eS
ou
ther
nH
emis
pher
e.V
alu
esin
Md
,P
,E
-ex
t.an
dB
are
atle
ast
twic
eh
igh
er(?
?)
or
atle
ast
twic
elo
wer
(--
)th
anin
S-e
xt.
;if
no
t(*
)
nd
no
dat
aa
Th
ed
om
inan
tp
hy
topla
nk
ton
gro
up
inar
eaS
-ex
t.is
PR
Ob
ut
NA
NO
or
CO
Cd
om
inan
ceis
also
det
ecte
dd
uri
ng
sum
mer
(Fig
.8a)
bT
he
do
min
ant
ph
yto
pla
nk
ton
pig
men
t-b
ased
size
clas
sin
area
S-e
xt.
isn
ano
ph
yto
pla
nk
ton
(Fig
.7d
)c
Insu
mm
erin
S-e
xt.
,[D
IP]
is2
–5
nM
(Lo
mas
etal
.2
01
0)
dD
IPtu
rno
ver
tim
eis
5h
(Co
tner
etal
.1
99
7).
Th
eli
mit
ing
nu
trie
nt
isP
(Wu
etal
.2
00
0)
eF
rom
enri
chm
ent
exp
erim
ents
fB
ut
unid
enti
fied
pix
els
cover
about
hal
fof
the
inves
tigat
edar
ea
Biogeochemistry
123
with Vallina and Simo (2007) and Miles et al. (2009),
k, the diffuse attenuation coefficient, can be deter-
mined from underwater PAR measurements (i.e.
k = 4.605/Ze, where Ze is the euphotic depth, at which
99% of the underwater PAR attenuates). Integrating
Eq. 3 over the MLD, with k now constant with depth,
leads to:
SRD ¼ ð1� aÞI #kMLD
1f � expð�kMLDÞg ð4Þ
Using this expression, the SRD values from the data
sets considered in this study can be compared as if they
were estimated similarly, despite their different origin.
In Table 5, some details are given on the data sets used
for calculating SRD from Eq. 4. The cruises PROS-
OPE, BIOSOPE and POMME were carried out in the
Mediterranean Sea, in the South Pacific gyre and in the
north-eastern Atlantic Ocean off Portugal, respec-
tively. BATS is the Bermuda Atlantic Time-Series
Study station. The DMS measurements were obtained
during those studies or as close as possible to those
locations (both geographically and temporally).
During PROSOPE, atmospheric PAR was mea-
sured on the ship instead of global incident irradiance
Table 3 Statistics concerning SRD and DMS/SRD relationship in the literature (local scale)
Location SRD (W m-2) N Slope r2 q References
Northwestern Med. Sea
Blanes Bay
10–290 15 0.028 0.94 0.75 Vallina and Simo (2007)
Bermuda Hydrostation S 5–230 33 0.017 0.81 0.89 Vallina and Simo (2007)
POMME cruisesa 0–90 232 0.051 0.19 0.74 Belviso and Caniaux (2009)
AMT cruisesb
In situ data 0–220 65 0.006 – 0.55 Miles et al. (2009)
Climatology 20–210 65 0.010 0.74
English Channel Station L4 0–55 33 0.13 0.25 – Archer et al. (2009)
Column 2 is the range of SRD values; N the number of samples used; Slope, r2 and q are respectively the slope, the coefficient of
determination and the Spearman rank correlation coefficient of the DMS/SRD regression. More details available in Table 4a Northeast Atlantic Ocean (39�N–44.5�N, 16.5�W–21�W)b Atlantic Ocean (45�N–40�S, roughly along 30�W)
Table 4 List of data sources and methods for calculating SRD in the literature
Location I(z) MLD k References
Blanes Bay Pyranometer; one exponential CTD; 0.2�C and 0 m PAR Vallina and Simo
(2007)
Bermuda
Hydrostation S
Pyranometer; one exponential CTD BATS; 0.1�C and 5 m Ze (PAR from BATS) Vallina and Simo
(2007)
Global Ocean 0.5 * IToA (computed from a
model); one exponential
BM04; 0.1�C and 5 m 0.06 m-1 Vallina and Simo
(2007)
POMME Pyranometer (300–2800 nm)
and satellite retrievals; one
and two exponentials
CTD; several criteria
including 0.1�C and 5 m
D1 and D2 from
Jerlov (1977)
Belviso and
Caniaux (2009)
AMT Pyranometer (300–3000 nm);
one exponential
CTD at 03 h LT; 0.1�C and
5 m
Ze at 11 h LT Miles et al. (2009)
English Channel
Station L4
ELDONET dosimeter;
(400–700 nm)
CTD; 0.8�C and 0 m PAR Archer et al. (2009)
Global Ocean ISCCP surface solar irradiance;
one exponential
BM04; 0.2�C and 10 m 0.06 m-1 Derevianko et al.
(2009)
I(z) refer to the solar irradiance decay in the water column; MLD stands for the mixed layer depth and k the light extinction
coefficient. BM04 refers to the de Boyer Montegut et al. (2004) climatology
Biogeochemistry
123
(motivating the use of the 0.43 factor for the other data
sets). For POMME and BATS, k was determined from
Chl profiles (unit mg m-3) using the following
expressions (Morel and Berthon 1989):
Ze ¼ 568:2
ZZe
0
ðChlÞdz
* +�0:746
when Ze\102 m
and
Ze ¼ 200
ZZe
0
ðChlÞdz
* +�0:293
when Ze [ 102 m
ð5ÞThen k and SRD were calculated for each CTD
by using the nearest time Ze value. During BIOSOPE
and PROSOPE, PAR profiles were only collected at
some stops of the ship and consequently Ze values
were calculated from these profiles and k and SRD
determined for each CTD by using the nearest Ze
value.
Standardization procedure for SRD estimates in the
UV domain. Monthly climatologies of the average
incident UV downward scalar irradiance in the mixed
layer, SRDðUVÞ (mol(photons) m-2 day-1), was
estimated for the studied regions using Eq. 6:
SRDðUVÞ � 1
MLD
ZMLD
0
Zk2
k1
Eodðk; 0�Þe�KdðkÞzdkdz
ð6Þ
whereEodðk; 0�Þ is the scalar downward irradiance just
below sea surface (mol(photons) m-2 day-1 nm-1),
and KdðkÞ is the average diffuse attenuation coefficient
of downwelling irradiance for the mixed layer (m-1).
SRDðUVÞwas computed for the combined UV-B/UV-
A region (290–400 nm).
The monthly climatologies of MLD, Eodðk; 0�Þ, and
KdðkÞ used in the calculations are described in detail in
Fichot and Miller (2010) along with a discussion about
uncertainties associated with assumptions made in the
modeling. Note that a spectral resolution of 2-nm was
used in the present work, in contrast to the 5-nm
spectral resolution used in Fichot and Miller (2010).
For each region studied, SRDðUVÞwas first calculated
for the individual elements of the spatial grid corre-
sponding to the region. Then, the spatial average and
standard deviation for each region was computed.
Note that SRD(UV) are expressed here in units of
mol(photons) m-2 day-1 whereas SRD in the PAR
domain are expressed in units of W m-2. There are
two main reasons for this discrepancy: 1) SRD(PAR)
are calculated here from broadband radiometric mea-
surements for which a conversion to quantum units
can lead to significant errors because some assumption
on the spectral characteristics of the incident irradi-
ance is required; 2) the absorption of a photon is the
fundamental process in photochemical reactions such
that SRD(UV) (modeled here from spectrally resolved
data) are best expressed in quantum units when
investigating the role of UV radiation on DMS
dynamics.
Biogeochemical variables relevant for the marine
sulfur cycle Phytoplankton group dominance, phyto-
plankton pigment-based size classes, photoprotection
index, mixed layer UV photon absorption rate by
CDOM together with dissolved inorganic phosphorus
(DIP) concentrations and turnover times (a measure of
phosphate limitation which can also be assessed from
enrichment experiments), were gathered from several
databases, from computed climatologies and from the
Table 5 List of input parameters, method and criteria used for the calculation of SRD
Experiment N MLD Surface irradiance k method
BATS (summers 1992–1994) 41 CTD 0.1� and 5 m Pyranometer Ze from Chl a profiles at some spots
PROSOPE (1999) 51 CTD 0.1� and 5 m PAR Ze from underwater PAR profiles at some spots
BIOSOPE (2004) 218 CTD 0.1� and 5 m Pyranometer Ze from underwater PAR profiles at some spots
POMME (2001) 232 CTD 0.1� and 5 m Pyranometer Ze from Chl a profiles at each CTD cast
N is the number of CTDs used, k the light extinction coefficient. However, only a fraction of the PROSOPE, BIOSOPE and POMME
data sets was used since our study is restricted temporally (summer months only i.e. cruise POMME 3 carried out in September
2001), and spatially (oligotrophic areas: site MIO during PROSOPE, and stations 9–12 during cruise BIOSOPE)
Biogeochemistry
123
literature (Table 2 and associated references). The first
three variables are relevant for the DMS cycle because
algal composition and the physiological conditions of
the algal cells affect DMSP and DMS production
(Stefels et al. 2007). Light and nutrient stresses are also
potentially important factors controlling bacterial
consumption of DMSP and DMS (Polimene et al. this
issue). Concerning the UV absorption by CDOM, a
potential source of free radicals likely involved in the
photo-oxidation of DMS (Toole and Siegel 2004),
monthly climatologies of the average UV photon
absorption rate in the mixed layer, NMLDðUVÞ(mol(photons) m-2 day-1), were estimated for the
studied regions using Eq. 7:
NMLDðUVÞ
� 1
MLD
ZMLD
0
Zk2
k1
Eodðk; 0�Þe�KdðkÞzagðkÞdkdz;ð7Þ
where agðkÞ is the average absorption coefficient of
CDOM for the mixed layer (m-1). NMLDðUVÞ was
computed for the combined UV-B/UV-A region. The
monthly climatologies of agðkÞ used in the calcula-
tions are described in detail in Fichot and Miller
(2010) along with a discussion of uncertainties
associated with assumptions made in the modeling.
A spectral resolution of 2-nm was used in the present
work, in contrast to the 5-nm spectral resolution used
in Fichot and Miller (2010). For each region studied,
NMLDðUVÞ was first calculated for the individual
elements of the spatial grid corresponding to the
region. Then, the spatial average and standard devi-
ation for each region was computed.
The ‘province based’ approach
The climatology of Kettle et al. (1999), recently
updated by Lana et al. (2011), uses interpolation/
extrapolation techniques that are applied to project the
discrete DMS surface concentration data onto a first
guess field based on Longhurst’s biogeographic prov-
inces. Further objective analysis allows the authors to
obtain the final monthly climatological maps. Figure 1
displays the Longhurst (1998) divisions of the sub-
tropical oceans into static biogeographic provinces.
Hence, we used monthly climatological data for DMS,
Chl, and POC averaged over the biogeographic
province wherein a given oligotrophic subtropical
zone resides (Table 1).
The ‘model based’ approach
The PISCES model
PISCES (acronym for ‘‘Pelagic Interaction Scheme for
Carbon and Ecosystem Studies’’) is a biogeochemical
model (Aumont and Bopp 2006), which simulates
marine biological production and describes the cycles
of carbon and of the main nutrients (phosphate, nitrate,
ammonium, iron and silicate). Two phytoplankton
functional groups (nanophytoplankton and diatoms
which are of central importance in the biogeochemical
cycles of sulfur and carbon, respectively) and two
zooplankton functional groups (mesozooplankton and
microzooplankton), as well as two detrital size classes
are also represented. PISCES is embedded within the
global general ocean circulation model NEMO (Madec
et al. 1998). PISCES was designed to be used over a
wide range of spatial and temporal scales. A complete
description of PISCES can be found in Aumont and
Bopp (2006) and a release of the model is available to
the community (http://www.nemo-ocean.eu/).
A new DMS module
The first prognostic module computing seawater DMS
concentrations and DMS air–sea fluxes embedded
within PISCES was developed by Bopp et al. (2008).
This model was used to explore the relationship
between iron fertilization and DMS production in the
Southern Ocean. For the purposes of this study, which
aims at investigating the dynamics of DMS in
oligotrophic regions, we introduced a new parameter-
ization into the DMS module. Most of the new
equations were taken from the PlankTOM5 model
(Vogt et al. 2010) which better captures the temporal
decoupling between chlorophyll and DMS in poorly
productive waters. This results in DMS concentrations
that are highest several weeks after Chl concentrations
have already peaked (e.g., Simo and Pedros-Alio
1999). Here, we only describe the equations and
parameters used in the DMS module.
The DMS module explicitly represents DMSP in
the particulate form (pDMSP) and dissolved DMS
according to the following equations:
Biogeochemistry
123
pDMSP ¼X
i
S
C
� �i
Ci; ð8Þ
oDMS
ot¼X
i
Gi þ Ei þMi½ � S
C
� �i
YS � k�DMSDMS
� kPARDMS � kNO3
DMS � DMS� kgDMS;
ð9Þ
where the subscript i represents the two phytoplankton
groups in PISCES, i.e. diatoms and nanophytoplank-
ton. The different terms of these two equations are
described below.
In Eq. 8, pDMSP is computed from the carbon
biomass of the two phytoplankton groups via group-
specific sulfur (DMSP)-to-carbon (S/C) ratios (Stefels
et al. 2007 and references therein). Recent studies have
highlighted large variability in S/C ratios within a
specific phytoplankton group as a function of envi-
ronmental factors (e.g. Sunda et al. 2002; Bucciarelli
and Sunda 2003). For nanophytoplankton and dia-
toms, we employ a variable S/C ratio that increases in
response to nutrient, light and temperature stress. We
use the equations for light (PAR), nutrient (Fe, DIP or
PO4) stress (S1, S2, S3, respectively), and temperature
(T) dependence as proposed by Vogt et al. (2010) to
simulate DMS in the PlankTOM5 model. The T
dependence simulates the function of DMS and DMSP
in cryoprotection (Karsten et al.1996). It has been set
to enhance the S/C ratio for low temperatures, so
mostly in polar waters. Hence, S/C ratio is described as
S
C
� �i
¼ S
C
� �i
min
þ S
C
� �i
var
� maxðS1; S2; S3; 0:3Þ � 2� fcorrð Þ
� 1þ 1
ðT þ 2:5Þ6
! ð10Þ
where
S1 ¼ PAR
PARmax
; ð11Þ
S2 ¼ KFe
Feþ KFe
; ð12Þ
S3 ¼ 0:7 � KPO4
PO4 þ KE4
; ð13Þ
In Eq. 10, SC
� �i
minand S
C
� �i
varare two parameters
representing the minimum and the variable part of
the S/C ratio for each phytoplankton. For example,
diatom cultures show that the S/C ratio of S.
costatum roughly doubles under nitrogen limitation
(Sunda et al. 2007). The S/C ratio of the diatom T.
oceanica roughly triples under Fe limitation (E.
Bucciarelli. pers. com.). That is why we have
introduced a variable part to the S/C ratio for
diatoms as for nanophytoplankton (Bopp et al.
2008). The S/C ratio is assumed to adapt instanta-
neously to the local conditions as observations
summarized in the review of Stefels et al. (2007)
suggest. In Eq. 10, fcorr is an adjustable correction
factor to tune global pDMSP concentration, currently
set to 0.25 (Vogt et al. 2010). As the UV radiation level
is not explicitly computed in PISCES, we assume that
high PAR is associated with high UV levels in order to
compute the S1 term in Eq. 11. In Eqs. 12 and 13 KN
are the phytoplankton half-saturation constants for
nutrients (Fe and PO4). The set of parameters used are
summarized in Table 6. Hence, S/C can vary between
0.015 and 0.030 for nanophytoplankton, and between
0.0018 and 0.0036 for diatoms.
In Eq. 9, pDMSP is released to seawater as
dissolved DMSP (dDMSP) via three different pro-
cesses: 1) grazing by zooplankton (Gi), 2) exudation
by phytoplankton (Ei), and 3) phytoplankton cell
lysis (Mi). These terms are modeled as per Lefevre
et al. (2002): in short, all pDMSP grazed by
zooplankton is released to seawater, the exudation
of dDMSP is similar to that of dissolved organic
matter (DOM), and the cell lysis is computed from
phytoplankton mortality. The amount of DMSP
released by these processes is proportional to the
S/C cell ratio. Ron Kiene’s main argument in 2000
was that dDMSP itself could be in short supply and
so figure into the determination of DMS yields as
dDMSP is broken down. Bacteria in need of sulfur
would keep the sulfur-containing molecule, which
then enters into DMSP catabolism via a demethyl-
ation pathway (Kiene et al. 2000). This core
reasoning does not appear in our DMS module
because we assume that, once released into water,
dDMSP is instantaneously transformed into DMS by
bacteria. We also assume that the dDMSP-to-DMS
yield (the percentage of DMSP cleaved by DMSP
lyases) is linked to bacterial activity. For instance, if
bacterial activity is low due to nutrient limitation,
the demethylation pathway is minor and DMS yield
increases. The dDMSP-to-DMS yield varies both
Biogeochemistry
123
spatially and temporally in a very large range (Simo
and Pedros-Alio 1999). Vezina (2004) reported
dDMSP-to-DMS yields in the range 18–68%. The
following parameterization was adopted to estimate
dDMSP-to-DMS yields (YS):
YS ¼ YS min þ YS var � 1� LBlim
� �; ð14Þ
LBlim ¼ min LB
PO4; LB
Fe; LBDOC
� �; ð15Þ
where
LBx ¼ x½ �= Kx þ x½ �ð Þ;
In this set of equations, LBlim is the bacterial nutrient
stress factor, and YS min and YS var are the minimum
and the variable part of the dDMSP-to-DMS yield,
respectively. In short, the modeled dDMSP-to-DMS
yield increases from 40% to 60% as a function of the
bacterial nutrient stress.
In seawater, DMS is lost due to bacterial
(k�DMSDMS) and photochemical (kPARDMS � kNO3
DMS
�DMS) processes, and ventilation to the atmosphere
(kgDMS). The bacterial consumption of DMS is
parameterized as a function of the bacterial biomass
(B) and activity:
Table 6 Model coefficients related to the DMS cycle with their standard values in PISCES, and references
Parameter Unit Value Description References
a day-1 0.66 Growth rate at 0�C for nanophyto.,
diatoms, and bacteria
Aumont and Bopp
(2006)
b – 1.066 Temperature sensitivity of growth Aumont and Bopp
(2006)
KPO4lM P 0.001/0.004/
0.0008
PO4 half-saturation constants: nano/diatoms/
bacteria
Aumont and Bopp
(2006)
KNH4lM N 0.013/0.065/
0.013
NH4 half-saturation constants: nano/diatoms/
bacteria
Aumont and Bopp
(2006)
KNO3lM N 0.26/1.3 NO3 half-saturation constants: nano/diatoms Aumont and Bopp
(2006)
KSi lM Si 2 Si half-saturation constant for diatoms Aumont and Bopp
(2006)
KFe nM Fe 0.02/0.1/0.02 Minimum Fe half-saturation constant for:
nano/diatoms/bacteria
Aumont and Bopp
(2006)
KDOC lM C 5 DOC half-saturation constant for bacteria Vogt et al. (2010)
KDMS-bac nM S 1.25 Half-saturation constant of DMS loss to bacterial
consumption
Archer et al. (2002)
(S/C)min-diato mol S (mol C)-1 0.0018 Minimum S-to-C ratio of diatoms Stefels et al. (2007)
(S/C)var-diato mol S (mol C)-1 0.0018 Variable S-to-C ratio of diatoms This work
(S/C)min-nano mol S (mol C)-1 0.015 Minimum S-to-C ratio of nanophytoplankton Stefels et al. (2007)
(S/C)var-nano mol S (mol C)-1 0.015 Variable S-to-C ratio of nanophytoplankton This work
YS min – 0.4 Minimum DMSP-to-DMS conversion yield Vezina (2004)
YS var – 0.2 Variable DMSP-to-DMS conversion yield This work
kDMSB day-1 0.2 DMS loss rate by bacterial activity Kieber et al. (1996)
PARmax W m-2 80 Maximal PAR Vogt et al. (2010)
f3 day-1 0.25 Maximum rate of photolysis of DMS to DOC Archer et al. (2004)
h1 W m-2 5 Michaelis–Menten PAR limitation value Archer et al. (2004)
Adapted from Bopp et al. (2008)
Biogeochemistry
123
k�DMS ¼ kBDMS � LB
lim�DMS � LBPAR � bT � BDMS � B
ð16Þ
LBlim�DMS ¼ min LB
lim;DMS
DMSþ KDMS�bac
� �ð17Þ
LBPAR¼min 1:;max 0:66;1:� PAR
PARmax
� �6
þ0:18
! !
ð18ÞThe parameterization in Eq. 16 is similar to the
formulation adopted in PISCES to compute the
remineralization rate of DOM (see Eq. 21 in Aumont
and Bopp 2006). The rate constant kBDMS for bacterial
consumption of DMS is set to 0.2 day-1, which is in
the low end of the range estimated by Kieber et al.
(1996) for the equatorial Pacific (0.2–0.6 day-1). The
limitation of bacterial activity LBlim�DMS � LB
PAR
� �was
taken from Vogt et al. (2010) (their Eqs. 10 and 11).
The temperature dependence, bT, is based on the
power law of Eppley (1972) for phytoplankton. BDMS
is a bacterial DMS coefficient currently set to 0.33
(Vogt et al. 2010). While PISCES does not explicitly
model bacterial biomass, it is estimated as a linear
combination of the zooplankton biomass terms
(B = 0.7 (Z ? 2 M) min(1, 120/z)) where Z and M
are the micro- and meso-zooplankton concentrations,
respectively, and z is the depth. This relationship has
been constructed from a version of PISCES that
included an explicit description of the bacterial
biomass (see Aumont and Bopp 2006).
The photochemical loss rate of DMS is a function of
PAR and nitrate (NO3) concentration. For the DMS
photolysis associated with PAR, we adopted the
parametrization proposed by Archer et al. 2004 (their
Eq. 12). Hence,
kPARDMS ¼ f3 �
PAR
PARþ h1
ð19Þ
where f3 is the maximum rate of photolysis of DMS to
dissolved organic carbon (DOC). We also include the
impact of NO3 on the photochemical loss processes
affecting DMS as suggested by Bouillon and Miller
(2004). This results in the following empirical
parameterization:
kNO3
DMS ¼ maxðLDMSNO3
; 0:01Þ ð20Þ
where
LDMSNO3¼ NO3½ �= KDMS
NO3þ NO3½ �
� �
where in the Michaelis–Menten function for NO3, the
half saturation constant KDMSNO3
� �= 10 lM.
For ventilation to the atmosphere, the gas exchange
coefficient, kg, is computed using the relationship as per
Wanninkhof (1992) and the Schmidt number for DMS
calculated following the formula given by Saltzman
et al. (1993). At present, estimates for kg vary by a factor
of 2, and the uncertainty increases with increasing wind
speed (Vogt and Liss 2009; Elliott 2009).
Experimental design
Here we used the global configuration of NEMO–
PISCES with a resolution of 2� 9 0.5 to 2� and 31
vertical levels forced by an ocean circulation gener-
ated from climatological atmospheric forcings (Au-
mont et al. 2008). Following a spin up of many
thousands of years we conducted short term sensitivity
experiments for less than 5 years in order to capture
the instantaneous response of the ocean’s DMS system
to change, rather than have the response complicated
by the lateral and downstream effects that would be
present in longer term integrations when nutrient
limitation is perturbed (e.g., Tagliabue et al. 2008). In
this way, we capture the direct effect of changes to
bacterial nutrient limitation on DMS cycling, rather
than the impacts associated with the modulation to
nutrient recycling and thus phytoplankton growth that
would also impact DMS cycling in longer term
sensitivity experiments.
Results
It is clear that the biogeographic provinces are
considerably less homogeneous in terms of Chl
(Fig. 1) or POC concentrations (data not shown) than
the cores of the subtropical gyres selected in the study
of Morel et al. (2010). Moreover, the core of the North
Atlantic subtropical gyre (South–Sargasso Sea)
belongs to two different provinces (NASW and
NATR). Hence, it is important to measure how the
conclusions of our study of the dynamics of DMS in
the most oligotrophic subtropical zones of the global
ocean are affected by the use of different DMS
datasets (i.e., discrete or climatological DMS data).
Biogeochemistry
123
Spatial variability of the ‘summer DMS paradox’
assessed from climatological data
Global features
Figure 3 examines the spatial variations in the inten-
sity of the ‘summer DMS paradox’ as calculated from
the normalized levels of climatological monthly DMS
concentrations to those of Chl (upper panel) or POC
(bottom panel) for the months of August and February.
Since POC exists in varying proportions to Chl, each
measure of the ‘summer DMS paradox’ exhibits
different trends in the Northern Hemisphere. There
are no major differences in maximum DMS-to-Chl
levels between the subtropical western North Pacific,
the subtropical North Atlantic and the Mediterranean
Sea. In terms of DMS-to-POC ratio, the subtropical
North Atlantic is characterized by maximum levels
about half of those observed in the two other basins.
Thus the maximum intensity of the ‘summer DMS
paradox’ is weaker in the Atlantic than in the other
oceanic basins and the Mediterranean Sea. In partic-
ular, both climatological data sets suggest that the
Indian South Subtropical Gyre is a better archetype for
the ‘summer DMS paradox’ than the Sargasso Sea. It
is also notable that a strong decoupling between DMS,
and Chl or POC concentrations, also occurs off the
Weddell Sea.
‘Province based’ approach: regional features
in the subtropical Longhurst’s biogeographic zones
The climatological monthly concentrations of DMS,
Chl and POC have been spatially averaged over 10
biogeographic provinces of Longhurst (1998), with
seven of them located in the Northern Hemisphere.
These monthly quantities are displayed in Fig. 4
together with the corresponding DMS-to-Chl and
DMS-to-POC ratios. For a detailed discussion of
features in monthly regional mean climatology of
DMS concentrations, the reader is referred to the paper
of Lana et al. (2011). In the low and temperate
latitudes of the Northern and Southern Hemispheres,
DMS concentrations follow strong seasonal patterns
with winter lows and summer highs except in the
provinces NATR, NPTW and WARM (Fig. 4a, b).
The WARM region follows a seasonal pattern with
much higher DMS concentrations in winter despite
little variation in Chl concentrations. This results in a
very unusual seasonal pattern in DMS-to-Chl and
DMS-to-POC ratios that is highly dissimilar to the
seasonal trends in the provinces NASW and MEDI
(Fig. 4g, i). Climatological data reveal that the ‘sum-
mer DMS paradox’ is of higher amplitude in MEDI
than elsewhere in the Northern Hemisphere and also
that it takes place in all subtropical provinces of the
Southern Hemisphere, but with variable intensity that
is highest in ISSG and lowest in SATL (Fig. 4g–j).
The ‘summer DMS paradox’ in NASW and SPSG
look very similar. Hence, according to our ‘province
based’ approach, the ‘summer DMS paradox’ is
widespread in subtropical oligotrophic waters.
‘Station based’ approach of ‘summer DMS
paradox’ and associated physical
and biogeochemical variables
Absolute and relative monthly DMS concentrations
The data used (1,427 DMS surface seawater concen-
tration measurements (Table 1) plotted as red dots in
Fig. 1) to investigate the monthly variations in DMS
concentrations in the most oligotrophic subtropical
zones of the global ocean (see the white boxes in Fig. 1)
are solely taken from the Global Surface Seawater
DMS Database (http://saga.pmel.noaa.gov/dms/). Data
were averaged in each zone according to the sampling
month (Fig. 5a) and normalized to climatological
monthly concentrations of Chl and POC (Fig. 5b–c).
Summer DMS mean levels in Md and S-ext. are similar.
If we use P as a comparison with S-ext., we find that the
levels across the subtropical North-Atlantic are mark-
edly different in spring and summer. In summer, there is
also a two- to threefold difference in DMS concentra-
tion between sites S-ext. and E-ext., and levels in site B
are about half of those in site E-ext. Hence, during the
summer the highest contrast is found between the sites
B and S-ext. It is worthy to note here that monthly
climatological DMS data averaged in the Longhurst’s
static biogeographic provinces (Fig. 4a, b) and monthly
DMS data averaged in the core of the subtropical oli-
gotrophic zones (Fig. 5a) provide distinct pictures. This
probably results from the interpolation/extrapolation/
substitution techniques applied to project the discrete
concentration data onto a first guess field based on
Longhurst’s biogeographic provinces. It highlights how
local dynamics can be ‘averaged out’ in climatology
construction at regional scales. The E-ext. zone is
Biogeochemistry
123
ambiguous as to whether a ‘summer DMS paradox’
exists there. Indeed, the DMS-to-Chl ratios calculated
in mid and late summer in E-ext. (February–March) and
S-ext. (August–September) are roughly similar
(Fig. 5b). In late spring and early summer (December–
January in the SH and June–July in the NH), the ratio in
E-ext. is about half that in S-ext. When the intensity of
the ‘summer DMS paradox’ is measured using DMS-
to-POC ratios, the results are less ambiguous. It appears
that the ‘summer DMS paradox’ is more important in
S-ext. and Md than in E-ext. and B.
Physical and chemical oceanic variables (MLD, kg,
SRD PAR and UV, NMLDðUVÞ)
Monthly values extracted or calculated from clima-
tologies are shown in Fig. 6. Here, we focus on
summer patterns.
DMS sea-to-air transfer velocities show little
differences between sites with the exception of area
H where wind speeds are generally higher than in the
other regions (Fig. 6b). Shallower mixed layer depths
are observed in S-ext., Md and P (\30 m, Fig. 6a).
Fig. 3 Climatological
monthly maps of DMS-to-
Chl (upper panel,mmol g-1) and DMS-to-
POC (bottom panel,lmol g-1) ratios for August
(NH summer) and February
(SH summer). Both ratios
measure the spatial
variability of the ‘summer
DMS paradox’ (Simo and
Pedros-Alio 1999)
Biogeochemistry
123
0.04
0.06
0.08
0.1
0.2C
hl (
mg
m-3
)
0
2
4
6
8
10
10.SATL37.SPSG23.ISSG
0
2
4
6
8
1018.NASE6.NASW7.NATR38.NPTE35.NPTW41.WARM16.MEDI
DM
S (
nM)
0.04
0.06
0.08
0.1
0.2
20
40
60
80
100
PO
C (
mg
m-3
)
20
40
60
80
100
N-Hemisphere
(a) (b)
(c) (d)
(g) (h)
(i) (j)
(e)
(f)
S-Hemisphere
0
25
50
75
100
125
150
DM
S/C
hl (
mm
ol g
-1)
0
25
50
75
100
125
150
0
50
100
150
200
DM
S/P
OC
(µm
ol g
-1)
0
50
100
150
200
Ja F M A M J Ju A S O N D Ju A S O N D Ja F M A M J
Fig. 4 Climatological
annual cycles of averaged a,
b DMS (Lana et al. 2011), c,
d corrected Chl (Morel and
Gentili 2009) and e, f POC
(Duforet-Gaurier et al.
2010) concentrations within
the 10 Longhurst static
biogeographic provinces
(see Table 1 for the meaning
of acronyms and reference
numbers). For each region
studied, the g, h DMS-to-
Chl and i, j DMS-to-POC
ratios were first calculated
for the 1� 9 1� pixels of the
spatial grid corresponding to
the region. Then, the spatial
average for each region was
computed. Both ratios
measure the amplitude of the
‘summer DMS paradox’
(Simo and Pedros-Alio
1999). The month scale
begins with January for the
Northern Hemisphere (leftpanels), and in July for the
Southern Hemisphere (rightpanels). Note that Chl and
POC concentrations are
plotted on a log scale
Biogeochemistry
123
The influence of MLDs on mixed layer SRDs is
obvious. In the UV domain, the highest SRDs are
found in Md (Fig. 6c). SRDs in S-ext. and E-ext. are
almost identical and are significantly higher than in B.
Lower SRDs are found in P, M and H.
In the PAR domain, mixed layer SRDs were
calculated from field measurements of irradiance,
SST and light attenuation that coincide geographically
and temporally with the DMS records. Therefore, we
provide discrete SRD data in the PAR domain with
units W m-2 as per Vallina and Simo (2007) together
with the number of vertical profiles used for the
determinations. SRDs were 75.1 ± 20.2 W m-2
(n = 41), 80.4 ± 6.7 W m-2 (n = 35), 91.9 ± 8.9
W m-2 (n = 13) and 43.8 ± 17.6 W m-2 (n = 74)
in S-ext., Md, E-ext. and P, respectively.
The UV photon absorption rate averaged over the
mixed layer was similar at all sites during summer,
with maximum rates observed at site B (Fig. 6d). The
reason for this similarity is essentially twofold. First,
all study sites are typically exposed to a comparable
level of incident UV irradiance during summer (see
Fig. 3A, C in Fichot and Miller 2010 for summer
values of incident irradiance in the Northern and
Southern Hemisphere, respectively). Second, the large
majority ([80%) of all incident UV photons is
generally absorbed within the mixed layer, even in
the subtropical gyres during summer when UV diffuse
attenuation is low and the mixed layer is shallow
(Fichot and Miller 2010).
Biological oceanic variables
Phytoplankton pigment-based size classes were used to
document the structure of the phytoplankton commu-
nity in each oligotrophic zone (Fig. 7). While total Chl
a (TChl a) is a universal proxy for all phytoplanktonic
organisms, accessory pigments measured by high
performance liquid chromatography (HPLC) are spe-
cific to phytoplankton groups. Their respective propor-
tion to TChl a can thus be used as a proxy for
community composition. Here, we used the pigment
grouping method recently improved by Uitz et al.
(2006). Seven pigments are used as biomarkers of
several phytoplankton taxa. Specific pigments and
classes are described in Uitz et al. (2006), see their
Table 2. These taxa are then grouped into three size
classes (micro-, nano-, and picophytoplankton (pro-
chlorophytes and cyanophytes)) according to the
0
20
40
60
80
100
120
DM
S/C
hl (
mm
ol g
-1)
0
2
4
6
8
(a)
(b)
(c)
PS-ext.HMMdBE-ext.
DM
S (
nM)
0
50
100
150
200
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
DM
S/P
OC
(µm
ol g
-1)
Fig. 5 Seasonal variations in monthly averaged DMS surface
seawater concentrations in the most oligotrophic subtropical
zones of the global ocean (a). The vertical bars correspond to
±1 standard deviation computed for each month and express the
combined spatial and year-to-year variability of the monthly
mean values inside each zone. Seasonal variations in monthly
averaged DMS surface seawater concentrations normalized to
monthly climatological Chl and POC concentrations (b, c). Note
that site ‘‘P’’ (off Portugal) is not within the core of an
oligotrophy zone and is used here for comparison
Biogeochemistry
123
average size of the cells. The fraction of each pigment-
based size class with respect to the total phytoplankton
biomass is calculated in each of the oligotrophic zones
on a monthly basis. Unfortunately, some areas still
remain poorly documented (e.g. E-ext., B and P;
Fig. 7). In area H, the total phytoplankton biomass was
generally higher than in the other regions and the
phytoplankton community was dominated by pic-
ophytoplankton year-round (Fig. 7e). In area S-ext.
during summer the dominant size class was nanophyto-
plankton. Picophytoplankton and nanophytoplankton
were the dominant classes in the oligotrophic system
Md. Summer data in E-ext. and B are lacking but the
late spring data show that picophytoplankton and
nanophytoplankton are the dominant classes.
The PHYSAT tool provides a way to trace the
dominant phytoplankton populations from space at an
unprecedented scale (Alvain et al. 2008). The PHY-
SAT algorithm, applied to the SeaWiFS daily L3-
binned GAC data archive, identifies the following
dominant phytoplankton groups in surface waters:
Prochloroccocus (PRO), Synechococcus (SYN), na-
noeukaryotes (NANO), Phaeocystis-like (PHAEO),
coccolithophores (COC), and diatoms (DIAT). Cli-
matological (1998–2006) monthly phytoplankton
group dominance (PGD) data were used to corroborate
the phytoplankton pigment-based size classes pre-
sented above. The main features in PGD in the
Northern Hemisphere in summer (August) were: (1)
SYN dominated in H and M, and (2) PRO and NANO
dominated in S-ext. (Fig. 8a). SYN dominated in the
areas E-ext., B and I in summer (February) in the
Southern Hemisphere (Fig. 8b).
The adaptation of phytoplankton communities to
extreme optical conditions (deep blue waters) can be
investigated through the use of the photoprotection
index PI = (Diadinoxanthin ? Diatoxanthin)/[TChl
a]. Diadinoxanthin and diatoxanthin, also measured
by HPLC, are non-photosynthetic pigments produced
by various phytoplankton taxa (mainly micro- and
0
50
100
150
200
250
NH
SH
Ja
Ju
F
A
M
S
A
O
M
N
J
D
Ju
Ja
A
F
S
M
O
A
N
M
D
J
BE-ext.PS-ext.MHMd
MLD
(m
)
0
5
10
15
20
25
NH
SH
Ja
Ju
F
A
M
S
A
O
M
N
J
D
Ju
Ja
A
F
S
M
O
A
N
M
D
J
DM
S s
ea-t
o-ai
r tr
ansf
er v
eloc
ity (
cm h
-1)
0
0.5
1
1.5
2
2.5
3
3.5
4
NH
SH
Ja
Ju
F
A
M
S
A
O
M
N
J
D
Ju
Ja
A
F
S
M
O
A
N
M
D
J
SR
D U
V A
&B
(m
ol p
hoto
ns.m
-2.d
-1)
0
0.5
1
1.5
2
2.5
3
NH
SH
Ja
Ju
F
A
M
S
A
O
M
N
J
D
Ju
Ja
A
F
S
M
O
A
N
M
D
J
UV
A&
B a
bsor
bed
(mol
pho
tons
.m-2
.d-1
)
(a) (b)
(d)(c)
Fig. 6 Monthly
climatological values of
a the depth of the mixed
layer, b the DMS sea-to-air
transfer velocity of DMS,
c the average incident UV
downward scalar irradiance
in the mixed layer (SRD
UV), and d the average UV
photon absorption rate in the
mixed layer, averaged in
each of the selected
oligotrophic zones
Biogeochemistry
123
nanophytoplanktion). They act to prevent photoinhi-
bition and reactive oxygen species production. Hence,
PI rather measures the light adaptation of micro- and
nanophytoplankton. The oligotrophic zones showing
higher values of photoprotection index are S-ext., Md
and E-ext. (Fig. 7b).
Dissolved inorganic phosphorus (DIP) concentra-
tions and turnover times (a measure of phosphate
limitation), were gathered from the literature
(Table 2). In S-ext. during summer, DIP concentra-
tions in surface waters are 2–5 nM (Lomas et al.
2010). They are also in the low nM levels in Md
where the depth of the 50 nM isoline is deeper than
150 m (Van Wambeke et al. 2002). In subtropical
and intertropical Atlantic surface waters, DIP levels
at 25 m are considerably higher (i.e. tens to hundreds
of nM; Mather et al. 2008). In E-ext., DIP levels are
over 200 nM (Moutin et al. 2008). Areas S-ext. and
Md display extremely short turnover times in DIP
(5 h and 1.2 ± 0.5 h; Moutin et al. 2008, respec-
tively) as compared to E-ext. and the north-eastern
Atlantic area P (5589 ± 1472 and 168 ± 110 h;
Moutin et al. 2008, respectively). Nutrient limitation
has also been assessed from enrichment experiments.
The north-western Atlantic and the eastern Mediter-
ranean Sea are areas where phytoplankton and
bacterial growth is limited by phosphate availability
(Wu et al. 2000; Van Wambeke et al. 2002) while
nitrogen is the limiting nutrient in area E-ext. (Van
Wambeke et al. 2008). According to Mather et al.
(2008), the South Atlantic subtropical gyre is not
phosphate-limited. These authors highlighted the
asymmetry of dissolved organic phosphorus (DOP)
cycling in the North and South Atlantic Ocean
subtropical gyres and interpreted it as a consequence
of enhanced nitrogen fixation in the North Atlantic
Ocean, which forces the system towards phosphorus
limitation.
‘Model based’ approach to DMS dynamics
Data obtained from satellite imagery by SeaWiFS
(reprocessed by Morel et al. 2010) offer the means to
validate the surface distribution of chlorophyll as
0
0.02
0.04
0.06
0.08
0.1
0.12
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
Mic
ro-C
hla
(mg
m-3
)
0
0.1
0.2
0.3
0.4
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
Pho
topr
otec
tion
Inde
x(D
D +
DT
)/T
Chl
a
0
0.05
0.1
0.15
0.2
0.25
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
BE-ext.PS-ext.HMd
TC
hla
(mg
m-3
)
0
0.02
0.04
0.06
0.08
0.1
0.12
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
Nan
o-C
hla
(mg
m-3
)
0
0.02
0.04
0.06
0.08
0.1
0.12
NHSH
JaJu
FA
MS
AO
MN
JD
JuJa
AF
SM
OA
NM
DJ
Pic
o-C
hla
(mg
m-3
)
(a) (b)
(c) (d) (e)
Fig. 7 Total phytoplankton biomass (a), light adaptation of phytoplankton (b) and phytoplankton pigment-based size classes (c micro-,
d nano- and e picophytoplankton) in each oligotrophic zone investigated
Biogeochemistry
123
predicted by the PISCES model. Figure 9 shows the
yearly mean, climatological, satellite-derived (a,
SeaWiFS) and modelled distributions (b, PISCES) of
chlorophyll at the surface. The satellite provides
average values for the upper 10–20 m of the ocean
that are directly comparable to the model output. The
main observed patterns are reproduced by the model.
Very low concentrations of Chl are found in the
oligotrophic subtropical gyres. They drop below
0.05 mg Chl m-3 both in the model and in the
observations. In the equatorial Pacific Ocean the
predicted chlorophyll concentrations are slightly
overestimated. The patterns observed in the latitude-
time (Hovmøller) maps confirms that the model is able
to closely simulate the spatial and seasonal variations
in Chl (Fig. 9c, d). In terms of DMS, the model
simulation and the annual climatological map con-
structed from Lana et al. (2011) diverge mostly in how
they represent DMS structures at mid- to high-
latitudes (Fig. 9e, f). Moreover, the climatology of
Lana et al. (2011) suggests zonal patterns with DMS
summer maxima coinciding with Chl maxima at high
SH latitudes (compare Fig. 9g and c). The coincidence
with Chl is reduced at mid- to high-NH latitudes. The
model produces zonal DMS fields that do not follow
closely those of Chl almost everywhere (compare
Fig. 9h and d). Moreover, at mid- to high-latitudes,
simulations and climatological data of Lana et al.
(2011) differ markedly (compare Fig. 9g and h). The
results of the sensitivity experiment in the subtropical
oligotrophic zones will be presented in the discussion.
Discussion
In their paper dated 2008, Vila-Costa et al. proposed
an explanation of the links from SRD to the biogeo-
chemical cycle of DMS: ‘‘In highly irradiated waters
typical of summer conditions, stronger stratification of
the water column favours high DMSP-producing
phytoplankton, and high UV irradiances in shallow
mixed layers generate high UV radiation doses for
plankton that may inhibit the consumption of DMSP
and DMS by bacteria and enhance the anti-oxidative
responses of phytoplankton, all resulting in higher
DMS production than consumption.’’ Our work aims
Fig. 8 Dominant
phytoplankton groups for
August (a) and February
(b) from PHYSAT
climatological composites
(1998–2006, Alvain et al.
2008). Dominant
phytoplankton groups are:
Nanoeucaryotes (NANO),
Prochloroccocus (PRO),
Synechococcus (SYN),
diatoms (DIAT),
Phaeocystis-like (PHAEO)
and coccolithophores
(COC)
Biogeochemistry
123
Biogeochemistry
123
at testing this causal link in other oligotrophic
subtropical zones of the global ocean than the
Sargasso Sea which is often considered as an arche-
type for oligotrophic regimes but is certainly not the
bluest sea (Morel et al. 2010). Indeed, the Sargasso
Sea, as well as the Mediterranean Sea, are not as blue
as it could be expected from their low Chl level,
essentially because of a relative CDOM excess (Morel
and Gentili 2009; Morel et al. 2010). Based on the high
correlation between DMS and SRD at local scales
(Vallina and Simo 2007), we expected that the clearest
and the most oligotrophic waters of all the subtropical
gyres, i.e. the South Pacific Gyre (Morel et al. 2010),
would display the highest DMS surface levels. SRDs
in the PAR domain, for which a new method of
calculation based on field observations of all its
components is proposed here, are indeed higher than
elsewhere (Fig. 10). However, we find that DMS
concentrations are two- to threefold lower in the South
Pacific Gyre than in the Sargasso or the eastern
Mediterranean Seas. Moreover, DMS levels in E-ext.
and P are roughly similar whereas SRDs in P are about
twice as low as in E-ext. (see also data of Fig. 7 in
Morel et al. 2010 for a representation of the annual
cycle of SRD in the PAR domain in the zones E-ext.
and P). DMS concentrations and SRDs in the UV
domain are strongly correlated in zone S-ext.
(Fig. 11a). At high SRDs, there is a better agreement
in terms of DMS levels between the S-ext. and Md
zones than between the former zone and E-ext. The
linear relationship in the S-ext. zone is even more
significant when DMS levels normalized to Chl or
POC concentrations are plotted against SRD-UV
(Fig. 11b, c). However, few data points from other
subtropical zones fit in. Moreover, most data points lay
below those regression lines. This is a pattern already
noticed by Miles et al. (2009) in their Atlantic Ocean
study of the relationship between total solar radiation
dose and DMS (compare Fig. 11a (this work) with
Fig. 2a in Miles et al. 2009). Overall, our observations
provide evidence that the SRD in the PAR and UV
spectral domains do not control DMS spatial varia-
tions in oligotrophic waters of the global ocean. The
amplitude of the excess in DMS concentration relative
to that of Chl or POC is probably modulated by other
factors than the penetration of solar radiation in the
sea. In other words, the causal link between SRD and
DMS which relies on a series of laboratory and field
observations (Dacey et al.1998; Aumont et al. 2002;
Vallina and Simo 2007; Sunda et al. 2007) should be
revised to integrate the fact that DMS-to-SRD, DMS-
to-Chl, DMS-to-POC ratios and most likely DMS-to-
DMSP ratios, differ from one oligotrophic ecosystem
to another.
The influences of physico-chemical and biological
variables on summer concentrations of DMS were
investigated in the most oligotrophic subtropical zones
of the global ocean. Unfortunately, very few DMS
measurements have been made in the South Indian
gyre and in the Mariana and Hawaii Islands zones
during summer. Similarities and differences in terms
of physical, chemical and biological variables between
the S-ext. zone and other oligotrophic subtropical
zones (Md, P, E-ext. and B) are summarized in
Table 2. The S-ext. and Md zones display strong
similarities in absolute or relative DMS concentra-
tions, in physical properties and in biogeochemical
regime: surface waters are highly stratified and DIP is
0
1
2
3
4
5
0 20 40 60 80 100 120
DM
S (
nM
)
SRD (W m-2)
PE-ext.
S-ext.
Md
Fig. 10 Scatter diagram of averaged DMS sea surface concen-
trations in four selected zones plotted against SRD in the PAR
spectral domain. In situ data for all of the components (MLD,
irradiance and light attenuation) have been used to calculate
SRDs
Fig. 9 Comparison of field observations (left panels) and
computer simulations performed using the PISCES model
(right panels). The four upper panels (a–d) show maps of annual
mean Chl concentration (mg m-3) and latitude-time (Hovmøl-
ler) maps of Chl (Morel et al. 2010). The four lower panels(e–h) are for DMS concentrations (nM; Lana et al. 2011)
b
Biogeochemistry
123
depleted, nanophytoplankton make an important con-
tribution to the phytoplankton community and are
adapted to high irradiances, DIP turnover rates are
extremely short (h) and phosphorous is the limiting
nutrient for phytoplankton and bacterial growth. In the
P zone, a highly stratified but not phosphorous-
depleted area exhibiting DIP turnover rates of hun-
dreds of hours, nanophytoplankton makes a smaller
contribution to the phytoplankton community and
does not show high levels of photoprotection index,
probably as a result of lower surface irradiance and
SRD. The Southern Hemisphere zone E-ext. is con-
siderably less stratified than the former three zones and
is distinct in terms of phytoplankton composition and
nutrient regime. The dominant phytoplankton group is
SYN and the limiting nutrient of phytoplankton and
bacterial growth in early summer is nitrogen. The B
zone is clearly not phosphate limited (Mather et al.
2008). If we remove the site ‘‘P’’ from the analysis
(because of the lack of permanent oligotrophy there,
but one should remind that site S-ext. exhibits no
permanent oligotrophy too), none of the variables
taken individually can account for the differences in
absolute and relative concentrations of DMS. Indeed,
the Northern and Southern Hemisphere zones are
distinct in terms of (1) water stratification (higher in
the NH), (2) phytoplankton taxa (more nanophyto-
plankton in NH), and (3) nutrient regime because
P-limitation of phytoplankton and bacterial growth
characterizes the NH sites S-ext. and Md. It is
impossible to evaluate which one of the three variables
is dominant. However, the lower DMS levels mea-
sured in summer in site ‘‘P’’ than in sites ‘‘S-ext.’’ and
‘‘Md’’ cannot result from changes in water stratifica-
tion or phytoplankton taxa because both variables are
very similar. Conversely, the DIP concentrations and
turnover times are sufficiently different to suggest that
the nutrient regime is the determining variable. The
causal link between SRD and high DMS levels in
summer would be valid only in phosphorus-limited
oligotrophic waters.
Phosphorus is an obligate requirement for growth
of all organisms (Van Mooy et al. 2009). The authors
also demonstrated that phytoplankton have an advan-
tage over heterotrophic bacteria for growth in phos-
phorus-limited oligotrophic waters, even if primary
production remains close to its lowest values in such
environments. The fundamental biochemical mecha-
nism allowing phytoplankton growth in phosphorus-
limited environments would be the ability of phyto-
plankton to substitute membrane phospholipids by
sulphur- and nitrogen-containing lipids that are free of
0
20
40
60
80
100
120
y = 1.9 + 38.7x r2= 0.91
DM
S/C
hl (
mm
ol g
-1)
0
1
2
3
4
5
6
7PS-ext.H
MMdB
E-ext.
y = 0.74 + 1.24x r2= 0.85
DM
S (
nM)
0
50
100
150
200
0 0.5 1 1.5 2 2.5 3 3.5
y = 21.1 + 64.6x r2= 0.92
DM
S/P
OC
(µ
mol
g-1
)
SRD-UV (mol photons m-2 d-1)
(a)
(b)
(c)
Fig. 11 Scatter diagrams of a averaged DMS sea surface
concentrations and b, c averaged DMS concentrations normal-
ized to Chl or POC concentrations, respectively, plotted against
SRD in the UV spectral domain. Climatological data for all of
the components (MLD, irradiance and light attenuation) has
been used to calculate the SRDs. The regression lines (plotted in
blue) are for the ‘‘S-ext.’’ dataset. See Table 1 for the exact
locations and meaning of the identifiers
Biogeochemistry
123
phosphorus (Van Mooy et al. 2009). Consequently, in
phosphorus-limited environments, phytoplankton
would maintain slow growth while heterotrophic
bacterial growth would be much more severely limited
by P-depletion. This would result in higher DMS
production than consumption and favour a build-up in
DMS. The Northern Atlantic subtropical gyre and the
Mediterranean Sea, which are subject to iron deposi-
tion from Saharan dust, are depleted in the nutrient
phosphate, possibly as a result of iron-enhanced
nitrogen fixation (Wu et al. 2000).
In this study, we use the PISCES ocean biogeo-
chemical model to explore the relationship between
phosphorus limitation and DMS production in oligo-
trophic subtropical waters. It is worth reminding here
how the biogeochemical model deals with the limita-
tion of phytoplankton and bacterial growth. PISCES
parameterizes the limitation of phytoplankton and
bacterial growth by nutrients under a Monod frame-
work using Michaelis–Menten kinetics, whereby the
half saturation constant for growth is used to calculate
the proportion of maximum growth permitted by each
nutrient in turn. The overall growth rate is then the
product of either the light limitation term (for phyto-
plankton) or the DOC limitation term (for bacteria)
and the limitation term for the most limiting nutrient.
For bacteria Fe, PO4 and NH4, whereas for phyto-
plankton NO3, NH4, PO4, Fe and (for diatoms)
Si(OH)4 are considered growth limiting nutrients. N2
fixation is included in PISCES as a process that is
assumed to result in the net production of reduced N
under certain conditions, in short it is enhanced under
high light, high temperature and low NO3 conditions
and its rate is further regulated by Fe or PO4 limitation.
The switch from phosphorus limited to non-limited
conditions was solely applied in the DMS module.
Variable sulfur-to-carbon ratios in phytoplankton,
variable transfer efficiency from DMS precursor to
DMS, and variable bacterial DMS-consumption rate
are taken into account. This ‘model based’ approach
shows that the DMS-to-Chl ratio can be heavily
impacted by phosphorus limitation in the North
Pacific, North Atlantic and most sectors of the Indian
Ocean, as the comparison of Fig. 12b, d and a, c
shows. The impact of phosphorus limitation extends in
the South Atlantic along the South American coast but
not in the core of the South Atlantic subtropical gyre
(Fig. 12c). There is no sensitivity to phosphorus
limitation in the core of the South Pacific gyre
(compare Fig. 12d with c). Both the S/C ratio of
nanophytoplankton and the DMSP-to-DMS conver-
sion yield are markedly affected (up to *-25%)
when phosphorus non-limited conditions are used
(Table 7). The nutrient stress factor of bacteria, which
also controls bacterial removal of DMS (Eqs. 16, 17),
is also heavily impacted by phosphorus limitation.
MLD integrated monthly budgets (lmol m-2
month-1) for August (Sargasso Sea) and February
(South Pacific and South Atlantic gyres; see Fig. 12
for exact locations) drawn for the standard run and the
sensitivity test are shown in Fig. 13. The DMS budget
in the South pacific gyre remains unaffected by
phosphorus limitation. In both cases, the removal of
DMS by bacteria accounts for 90% of the total
production (i.e. 90% of production is processed by
bacteria). When phosphorus limitation is released in
the Sargasso Sea, DMS production decreases from
160 to 100 lmol m-2 month-1 and DMS removal
increases from 56 to 72 lmol m-2 month-1. As a
consequence, only 35% of DMS production is pro-
cessed by bacteria under phosphorus-limited condi-
tions, whereas * 75% of DMS production is
processed by bacteria under phosphorus-replete con-
ditions. In the South Atlantic gyre, the DMS budget is
affected by P limitation but more marginally than in
the North Atlantic gyre. Finally, model results show
that the Indian south subtropical gyre is the sole SH
area where ‘strong summer paradox’ takes place
throughout the whole basin. This is consistent with the
climatological analysis but remains to be confirmed
from field observations in the core of this oligotrophic
zone.
Summary and conclusions
The most oligotrophic subtropical zones of the global
ocean have been compared with respect to their DMS,
Chl and POC contents by using two different
approaches to the ‘summer DMS paradox’. The first
approach documented the seasonal variations of
absolute and relative concentrations of DMS in the
corresponding Longhurst’s static biogeographic prov-
inces, using values extracted from a recently updated
monthly climatology of DMS and monthly satellite
composites of Chl and POC concentrations. DMS-to-
Chl and DMS-to-POC ratios exhibited strong seasonal
cycles in the North and South Atlantic gyres, the South
Biogeochemistry
123
Pacific gyre, the Indian south subtropical gyre as well
as in the Mediterranean Sea, with maxima in summer
and minima in winter. Thus, it appears that a ‘summer
DMS paradox’ is a common feature in most oligo-
trophic subtropical biogeographic provinces. The
second approach documented the monthly variations
of absolute and relative concentrations of DMS in the
cores of the most oligotrophic subtropical zones of the
global ocean, using discrete values extracted from the
Global Surface Seawater DMS Database and monthly
satellite composites of Chl and POC concentrations
since supporting Chl and POC data are not always
available from this database. The sites located in the
Northwestern Atlantic (S-ext.) and the eastern basin of
the Mediterranean Sea (Md) exhibited two- to three-
fold higher DMS levels in summer than in other NH
and SH locations. The surface radiation dose (SRD)
was calculated both in the PAR and UV spectral
domains and its influence on absolute and relative
concentrations of DMS was investigated through
monthly and spatial variations. The influence of
DMS/Chlmmol g-1
Standard run No P-limitation A
ug
ust
Feb
ruary
40°N
20°N
0°
40°N
20°N
0°
0°
20°S
40°S
0°
20°S
40°S
160°W 120°W 80°W 40°W 0° 40°E 80°E 120°E 160°E 160°W 120°W 80°W 40°W 0° 40°E 80°E 120°E 160°E
160°W 120°W 80°W 40°W 0° 40°E 80°E 120°E 160°E 160°W 120°W 80°W 40°W 0° 40°E 80°E 120°E 160°E
0 5 10 15 20 40 60 80 100 120 140 160 180
Fig. 12 Comparison of computer simulations performed using
the PISCES model: with P limitation (standard run; left panels(a, c)), without P-limitation (right panels (b, d)). Results are
shown for the summer months (August in NH and February in
SH; top and bottom rows, respectively). Black boxes correspond
to selected locations where DMS budgets were calculated (see
text). A short run was performed to capture the immediate
response of the DMS module to these changes rather than the
longer term feedbacks mediated by ocean circulation. The
simulation for the fourth year is presented here
Table 7 MLD averaged values of S/C ratio (units is mol -
mol-1) for nanophytoplankton, DMSP-to-DMS conversion
yield (YS) and bacterial nutrients stress factor (LlimB ) estimated
using PISCES at selected locations: Sargasso Sea close to the
BATS station, in the core of the South Pacific subtropical gyre
and in the South Atlantic gyre (see black boxes in Fig. 12)
Standard run (MLD
averaged)
No P-limitation
(%)
Sargasso Sea
(BATS)
S/Cnano = 0.030 -17
YS = 0.59 -24
LlimB = 0.008 ?8600
South Pacific
gyre
S/Cnano = 0.025 0
YS = 0.47 0
LlimB = 0.67 0
South Atlantic
gyre
S/Cnano = 0.030 -5
YS = 0.59 -5
LlimB = 0.57 ?22
Relative changes of S/Cnano, YS and LlimB averaged values for
the sensitivity experiment (no P-limitation) relative to the
standard run data are expressed in %. LlimB affects both YS and
the bacterial consumption of DMS (k�DMS, Eqs. 16 and 17)
Biogeochemistry
123
SRD on DMS was much less important outside the
S-ext. and Md zones. The differences in absolute and
relative concentrations of DMS between five oligo-
trophic sites were better accounted for by their nutrient
regimes than by stratification or phytoplankton taxa. In
other words, it appears that a strong ‘summer DMS
paradox’ preferentially takes place in phosphorus-
limited zones. A sensitivity experiment performed
using a new version of the PISCES 3D model
demonstrated the existence of a potential causal link
between P limitation and the ‘summer DMS paradox’.
Acknowledgments We are grateful to J.E. Johnson (JISAO,
University of Washington) and T.S. Bates (PMEL, NOAA) for
the maintenance and personal involvement in the Global Surface
Seawater DMS Database. The authors also wish to thank each of
the participants in the joint initiative of the SOLAS Integration
project and the EU project COST Action 735 for providing an
updated DMS climatology, with special thanks to the lead
author A. Lana. A. Longhurst kindly provided the biogeographic
provinces data. We would like to thank A. Morel and B. Gentili
for sharing ocean color data. We are grateful to Dr. Scott Elliott
and an anonymous reviewer for improvements to an earlier
version of this manuscript. The model work was performed
using HPC resources from GENCI-IDRIS (Grant 2009-010040
LSCE contribution 4662).
References
Alvain S, Moulin C, Dandonneau Y, Loisel H (2008) Seasonal
distribution and succession of dominant phytoplankton
groups in the global ocean: a satellite view. Glob Biogeo-
chem Cycles. doi:10.1029/2007GB003154
Andreae MO (1990) Ocean–atmosphere interactions in the
global biogeochemical sulfur cycle. Mar Chem 30:1–29
Archer SD, Smith GC, Nightingale PD, Widdicombe CE, Tarran
GA, Rees AP, Burkill PH (2002) Dynamics of particulate
dimethylsulphoniopropionate during a Lagrangian experi-
ment in the northern North Sea. Deep-Sea Res Part II 49:
2979–2999
Archer SD, Gilbert FJ, Allen JI, Blackford J, Nightingale PD
(2004) Modelling of the seasonal patterns of dimethylsul-
phide production and fate during 1989 at a site in the North
Sea. Can J Fish Aquat Sci 61:765–787
Archer SD, Cummings DG, Llewellyn CA, Fishwick JR (2009)
Phytoplankton taxa, irradiance and nutrient availability
determine the seasonal cycle of DMSP in temperate shelf
seas. Mar Ecol Prog Ser 394:111–124
Aumont O, Bopp L (2006) Globalizing results from ocean in situ
fertilization experiments. Glob Biogeochem Cycles. doi:
10.1029/2005GB002591
Aumont O, Belviso S, Monfray P (2002) Dimethylsulfoniopro-
pionate (DMSP) and dimethylsulfide (DMS) sea surface
distributions simulated from a global threedimensional
ocean carbon cycle model. J Geophys Res Oceans. doi:
10.1029/1999JC000111
Aumont O, Bopp L, Schulz M (2008) What does temporal
variability in aeolian dust deposition contribute to sea-
surface iron and chlorophyll distributions? Geophys Res
Lett. doi:10.1029/2007GL031131
Belviso S, Caniaux G (2009) A new assessment in North
Atlantic waters of the relationship between DMS concen-
tration and the upper mixed layer solar radiation dose. Glob
Biogeochem Cycles. doi:10.1029/2008GB003382
Bopp L, Aumont O, Belviso S, Blain S (2008) Modeling the
effect of iron fertilization on dimethylsulfide emissions in
the Southern Ocean. Deep-Sea Res Part II 35:901–912
Bouillon RC, Miller WL (2004) Determination of apparent
quantum yield spectra of DMS photo-degradation in an in
situ iron-induced Northeast Pacific Ocean bloom. Geophys
Res Lett. doi:10.1029/2004GL019536
Bucciarelli E, Sunda WG (2003) Influence of CO2, nitrate,
phosphate, and silicate limitation on intracellular dime-
thylsulfoniopropionate in batch cultures of the coastal
diatom Thalassiosira pseudonana. Limnol Oceanogr 48:
2256–2265
Cotner JB, Ammerman JW, Peele ER, Bentzen E (1997)
Phosphorus-limited bacterioplankton growth in the Sar-
gasso Sea. Aquat Microb Ecol 13:141–149
Dacey JWH, Howse FA, Michaels AF, Wakeham SG (1998)
Temporal variability of dimethylsulfide and dimethylsul-
foniopropionate in the Sargasso Sea. Deep-Sea Res Part I
45:2085–2104
de Boyer Montegut C, Madec G, Fischer AS, Lazar A, Iudicone
D (2004) Mixed layer depth over the global ocean: an
examination of profile data and profile-based climatology.
J Geophys Res Oceans. doi:10.1029/2004JC002378
Derevianko GJ, Deutsch C, Hall A (2009) On the relationship
between ocean DMS and solar radiation. Geophys Res
Lett. doi:10.1029/2009GL039412
Duforet-Gaurier L, Loisel H, Dessailly D, Nordkvist K, Alvain S
(2010) Estimates of particulate organic carbon over the
Fig. 13 DMS budgets drawn in the 3 selected zones Sargasso
Sea-BATS (S_BATS), E-ext. and B (see Fig. 12 for printed
locations). The computer simulations were performed using the
PISCES model: with P limitation (standard run; top row),
without P limitation (bottom row). Production and removal
fluxes of DMS in units of lmol m-2 month-1 were integrated
vertically (MLD; 25, 43 and 50 m, S_BATS, E-ext. and B,
respectively) and averaged in each zone for August (S_BATS)
and February (E-ext. and B locations). (Pro) stands for DMS
production, (v) for ventilation to atmosphere, (p) for photode-
gradation and (b) for bacterial removal. The ratio of bacterial
removal to DMS production is expressed in % and plotted
between brackets below the bacterial removal fluxes
Biogeochemistry
123
euphotic depth from in situ measurements. Application to
satellite data over the global ocean. Deep-Sea Res Part I
57:351–367
Elliott S (2009) Dependence of DMS global sea-air flux distri-
bution on transfer velocity and concentration field type.
J Geophys Res Oceans. doi:10.1029/2008JG000710
Eppley RW (1972) Temperature and phytoplankton growth in
the sea. Fish Bull 70:1063–1085
Fichot CG, Miller WL (2010) An approach to quantify depth-
resolved marine photochemical fluxes using remote sens-
ing: application to carbon monoxide (CO) photoproduc-
tion. Remote Sens Environ 114:1363–1377
Fichot CG, Sathyendranath S, Miller WL (2008) SeaUV and
SeaUVc: algorithms for the retrieval of UV diffuse atten-
uation coefficients from ocean color. Remote Sens Environ
112(4):1584–1602
Jerlov NG (1974) A simple method for measuring quanta irra-
diance in the ocean, Rep. 24. Inst. Fysik Oceanografi,
Kobenhavens Universitet
Jerlov NG (1977) Classification of sea water in terms of quanta
irradiance. J Cons Int Explor Mer 37(3):281–287
Jitts HR, Morel A, Saijo Y (1976) The relation of oceanic pri-
mary production to available photosynthetic irradiance.
Aust J Mar Freshw Res 27:441–454
Kara AB, Rochford PA, Hurlburt HE (2003) Mixed layer depth
variability over the global ocean. J Geophys Res Oceans.
doi:10.1029/2000JC000736
Karsten U, Kuck K, Vogt C, Kirst GO (1996) Dimethylsulfo-
niopropionate production in phototrophic organisms and its
physiological function as a cryoprotectant. In: Kiene RP,
Visscher PT, Keller MD, Kirst GO (eds) Biological and
environmental chemistry of DMSP and related sulfonium
compounds. Plenum, New York, pp 143–153
Katsaros K, Mcmurdie L, Lind R, DeVault J (1985) Albedo of a
water surface, spectral variation, effects of atmospheric
transmittance, sun angle and wind speed. J Geophys Res
Oceans 90:7313–7321
Kettle AJ, Andreae MO, Amouroux D, Andreae TW, Bates TS,
Berresheim H, Bingemer H, Boniforti R, Curran MAJ,
DiTullio GR, Helas G, Jones GB, Keller MD, Kiene RP,
Leck C, Levasseur M, Malin G, Maspero M, Matrai P,
McTaggart AR, Mihalopoulos N, Nguyen BC, Novo A,
Putaud JP, Rapsomanikis S, Roberts G, Schebeske G,
Sharma S, Simo R, Staubes R, Turner S, Uher G (1999) A
global database of sea surface dimethylsulfide (DMS)
measurements and a procedure to predict sea surface DMS
as a function of latitude, longitude, and month. Glob Bio-
geochem Cycles 13:399–444
Kieber DJ, Jiao JF, Kiene RP, Bates TS (1996) Impact of
dimethylsulfide photochemistry on methyl sulfur cycling in
the equatorial Pacific Ocean. J Geophys Res Oceans
101:3715–3722
Kiene RP, Linn LJ, Bruton JA (2000) New and important roles
for DMSP in marine microbial communities. J Sea Res
43:209–224
Lana A, Bell TG, Simo R, Vallina SM, Ballabrera-Poy J, Kettle
AJ, Dachs J, Bopp L, Saltzman ES, Stefels J, Johnson JE,
Liss PS (2011) An updated climatology of surface dim-
ethlysulfide concentrations and emission fluxes in the
global ocean, Glob Biogeochem Cycles. doi:10.1029/2010
GB003850
Langley L, Leaitch WR, Lohmann U, Shantz NC, Worsnop DR
(2010) Contributions from DMS and ship emissions to
CCN observed over the summertime North Pacific. Atmos
Chem Phys 10:1287–1314
Lefevre M, Vezina A, Levasseur M, Dacey JWH (2002) A
model of dimethylsulphide dynamics for the subtropical
North Atlantic. Deep-Sea Res Part I 49:2221–2239
Lomas MW, Burke AL, Lomas DA, Bell DW, Shen C, Dyhrman
ST, Ammerman JW (2010) Sargasso Sea phosphorus bio-
geochemistry: an important role for dissolved organic
phosphorus (DOP). Biogeosciences 7:695–710
Longhurst A (1998) Ecological geography of the sea. Academic,
London
Madec G, Delecluse P, Imbard M, Levy C (1998) OPA8.1 ocean
general circulation model reference manual. Notes du pole
de modelisation de l’IPSL
Mather RL, Reynolds SE, Wolff GA, Williams RG, Torres-Val-
des S, Woodward ME, Landolfi A, Pan X, Sanders R, Ach-
terberg EP (2008) Phosphorus cycling in the North and South
Atlantic Ocean subtropical gyres. Nat Geosci 1:439–443
Miles CJ, Bell TG, Lenton TM (2009) Testing the relationship
between the solar radiation dose and surface DMS con-
centrations using in situ data. Biogeosciences 6:1927–1934
Morel A, Berthon JF (1989) Surface pigments, algal biomass
profiles, and potential production of the euphotic layer:
relationships reinvestigated in view of remote-sensing
applications. Limnol Oceanogr 34(8):1545–1562
Morel A, Gentili B (2009) The dissolved yellow substance and
the shades of blue in the Mediterranean Sea. Biogeo-
sciences 6:2625–2636
Morel A, Claustre H, Gentili B (2010) The most oligotrophic
subtropical zones of the global ocean: similarities and
differences in terms of chlorophyll and yellow substance.
Biogeosciences 7:3139–3151
Moutin T, Karl DM, Duhamel S, Rimmelin P, Raimbault P, Van
Mooy BAS, Claustre H (2008) Phosphate availability and
the ultimate control of new nitrogen input by nitrogen
fixation in the tropical Pacific Ocean. Biogeosciences 5:
95–109
Payne RE (1972) Albedo of the sea surface. J Atmos Sci
29:959–970
Polimene L, Archer SD, Butenschon M, Allen JI (this issue) A
mechanistic explanation of the Sargasso Sea DMS ‘‘sum-
mer paradox’’. Biogeochemistry
Ruggaber R, Dlugi RA, Nakajima T (1994) Modelling of radi-
ation quantities and photolysis frequencies in the tropo-
sphere. J Atmos Chem 18(2):171–210
Saltzman ES, King DB, Holmen K, Leck C (1993) Experimental
determination of the diffusion coefficient of dimethylsul-
fide in water. J Geophys Res Oceans 98:16481–16486
Simo R, Pedros-Alio C (1999) Role of vertical mixing in con-
trolling the oceanic production of dimethyl sulphide. Nat-
ure 402:396–399
Stefels J, Steinke M, Turner S, Malin G, Belviso S (2007)
Environmental constraints on the production and removal
of the climatically active gas dimethylsulphide (DMS) and
implications for ecosystem modelling. Biogeochemistry
83(1–3):245–275
Sunda W, Kieber DJ, Kiene RP, Huntsman S (2002) An anti-
oxidant function for DMSP and DMS in marine algae.
Nature 418:317–320
Biogeochemistry
123
Sunda WG, Hardison R, Kiene RP, Bucciarelli E, Harada H
(2007) The effect of nitrogen limitation on cellular DMSP
and DMS release in marine phytoplankton: climate feed-
back implications. Aquat Sci 69:341–351
Tagliabue A, Bopp L, Aumont O (2008) Ocean biogeochemistry
exhibits contrasting responses to a large scale reduction in
dust deposition. Biogeosciences 5:11–24
Todd JD, Rogers R, Li YG, Wexler M, Bond PL, Sun L, Curson
ARJ, Malin G, Steinke M, Johnston AWB (2007) Struc-
tural and regulatory genes required to make the gas dime-
thyl sulfide in bacteria. Science 315:666–669
Toole DA, Siegel DA (2004) Light-driven cycling of dimeth-
ylsulfide (DMS) in the Sargasso Sea: Closing the loop.
Geophys Res Lett. doi:10.1029/2004GL019581
Uitz J, Claustre H, Morel A, Hooker S (2006) Vertical distri-
bution of phytoplankton communities in open ocean: an
assessment based on surface chlorophyll. J Geophys Res
Oceans. doi:10.1029/2005JC003207
Vallina SM, Simo R (2007) Strong relationship between DMS
and the solar radiation dose over the global surface ocean.
Science 315:506–508
Van Mooy BAS, Fredricks HF, Pedler BE, Dyhrman ST, Karl
DM, Koblızek M, Lomas MW, Mincer TJ, Moore LR,
Moutin T, Rappe MS, Webb EA (2009) Phytoplankton in
the ocean use non-phosphorus lipids in response to phos-
phorus scarcity. Nature 458:69–72
Van Wambeke F, Christaki U, Giannakourou A, Moutin T,
Souvemerzoglou K (2002) Longitudinal and vertical trends
of bacterial limitation by phosphorus and carbon in the
Mediterranean Sea. Microbiol Ecol 43:119–133
Van Wambeke F, Bonnet S, Moutin T, Raimbault P, Alarcon G,
Guieu C (2008) Factors limiting heterotrophic bacterial
production in the southern Pacific Ocean. Biogeosciences
5:833–845
Vezina AF (2004) Ecosystem modelling of the cycling of
marine dimethylsulfide: a review of current approaches and
of the potential for extrapolation to global scales. Can J
Fish Aquat Sci 61:845–856
Vila-Costa M, Kiene RP, Simo R (2008) Seasonal variability of
the dynamics of dimethylated sulfur compounds in a
coastal northwest Mediterranean site. Limnol Oceanogr
53(1):198–211
Vogt M, Liss PS (2009) Dimethylsulfide and climate. In: Le
Quere C, Saltzman ES (eds) Surface ocean—lower atmo-
sphere processes. Geophysical Monograph Series, vol 198,
pp 197–232
Vogt M, Vallina SM, Buitenhuis ET, Bopp L, Le Quere C
(2010) Simulating dimethylsulphide seasonality with the
Dynamic Green Ocean Model PlankTOM5. J Geophys Res
Oceans. doi:10.1029/2009JC005529
Wanninkhof R (1992) Relationship between wind speed and gas
exchange over the ocean. J Geophys Res Oceans
97:7373–7382
Wu J, Sunda W, Boyle EA, Karl DM (2000) Phosphate deple-
tion in the Western Atlantic Ocean. Science 289:759–762
Biogeochemistry
123